In least-squares reverse time migration (LSRTM), which is typically an overdetermined linear inverse problem, the kinematic travel time error from an inaccurate migration velocity leads to migration image degradation and slow convergence speed. Extended least-squares reverse time migration (ELSRTM), with an extended imaging condition along the subsurface offset in LSRTM, can effectively mitigate sensitivity to the migration velocity accuracy by adding an extra dimension to the model space. However, for assembling migration images, large imaging condition operations that are proportional to the number of subsurface offset bins hinder the practical application of ELSRTM. To address this computational problem in ELSRTM, we developed an efficient ELSRTM method based on a modified excitation amplitude (ExA) imaging condition. Furthermore, our ELSRTM method can correctly represent the forward source wavefield by convolving the source wavelet with the ExA. With the forward source wavefield represented this way, Born-modeled data can be simulated efficiently. Thus, wavefield simulation for generating the forward source wavefield is implemented only once at the first iteration. Based on these computational advantages, our ELSRTM method becomes highly efficient. Synthetic data examples on a modified marmousi2 velocity model demonstrate that our ELSRTM method produces an accurate migration image efficiently even if an inaccurate migration velocity is used. Field data acquired from the Chukchi Sea of the Arctic Ocean are also used to verify the practicality of our ELSRTM method.